Re: [PATCH] bpf: fix overflow of bpf_jit_limit when PAGE_SIZE >= 64K

Michael Roth [off-list ref] writes:

Commit ede95a63b5 introduced a bpf_jit_limit tuneable to limit BPF
JIT allocations. At compile time it defaults to PAGE_SIZE * 40000,
and is adjusted again at init time if MODULES_VADDR is defined.

For ppc64 kernels, MODULES_VADDR isn't defined, so we're stuck with

But maybe it should be, I don't know why we don't define it.

the compile-time default at boot-time, which is 0x9c400000 when
using 64K page size. This overflows the signed 32-bit bpf_jit_limit
value:

  root@ubuntu:/tmp# cat /proc/sys/net/core/bpf_jit_limit
  -1673527296

and can cause various unexpected failures throughout the network
stack. In one case `strace dhclient eth0` reported:

  setsockopt(5, SOL_SOCKET, SO_ATTACH_FILTER, {len=11, filter=0x105dd27f8}, 16) = -1 ENOTSUPP (Unknown error 524)

and similar failures can be seen with tools like tcpdump. This doesn't
always reproduce however, and I'm not sure why. The more consistent
failure I've seen is an Ubuntu 18.04 KVM guest booted on a POWER9 host
would time out on systemd/netplan configuring a virtio-net NIC with no
noticeable errors in the logs.

Fix this by limiting the compile-time default for bpf_jit_limit to
INT_MAX.

INT_MAX is a lot more than (4k * 40000), so I guess I'm not clear on
whether we should be using PAGE_SIZE here at all. I guess each BPF
program uses at least one page is the thinking?

Thanks for tracking this down. For some reason none of my ~10 test boxes
have hit this, perhaps I don't have new enough userspace?

You don't mention why you needed to add BPF_MIN(), I assume because the
kernel version of min() has gotten too complicated to work here?

Daniel I assume you'll merge this via your tree?

cheers